High Throughput Targeted Gene Sequencing in 738 Myelodysplastic Syndromes Patients Reveals Novel Oncogenic Genes, Rare Driver Mutations and Complex Molecular Signatures with Potential Impact for Patient Diagnosis and Prognosis in the Clinic

Abstract LBA-5

Myelodysplastic syndromes (MDS) are a diverse group of chronic hematological malignancies characterised by dysplasia, ineffective hematopoiesis and a variable risk of progression to acute myeloid leukemia (AML). Recent screens of MDS genomes have identified a diverse range of mutations in genes implicated in RNA splicing, DNA modification, chromatin regulation and cell signalling. Whilst these findings improve our understanding in the molecular aetiology of this disease a detailed understanding of the specific genetic contributions, genotype-phenotype relationships and prognostic value of this heterogeneous genomic landscape is warranted.

To evaluate the genomic architecture of MDS in more detail, targeted sequencing analysis was performed for 111 genes implicated in myeloid malignancy across 738 MDS patients. Sequencing data were assessed for the presence of oncogenic mutations and detection of regional copy number changes. High-confidence oncogenic mutations (n = 1155) were identified in 41 genes. Of these 41 genes, only 4 (SF3B1, TET2, SRSF2, ASXL1) were mutated in >10% of patients with 15 genes mutated between 10% and 1% of patients and 22 in < 1%. Overall, oncogenic lesions are identified in 78% of MDS patients with 43% of the patients displaying two or three oncogenic events and approximately 10% of patients harbouring four or more oncogenic mutations. Amongst the less characterised genes, we confirm inactivating mutations in CUX1 and IRF1 in 2% and 0.5% of the patients respectively and provide strong evidence that these two genes are bona-fide rare tumour suppressors in MDS. We observe complex patterns of pairwise association between genes, including co-mutation and mutual-exclusivity. Of 325 possible pairwise associations, 47 driver gene mutations pairs were significant (FDR < 10%). Mutations in components of the U2-spliceosome machinery are characteristic of 51% of patients with MDS and whilst mutually exclusive with one another, spliceosome mutations show a strong tendency to co-occur with genes implicated in DNA methylation.

Of 24 genes mutated in >5 patients, 8 genes were associated with significantly worse leukemia-free survival if mutated and one gene (SF3B1) with a better leukemia-free survival replicating findings from previous studies. For the new MDS genes described, loss of function truncating mutations in CUX1 are associated with a poor prognosis (p=0.02). Additionally, clinical outcome analysis identifies functional non-canonical mutations within well-characterised driver genes. We find that there is an inverse correlation of leukemia-free survival with the total number of oncogenic lesions observed (p<0.0001). Patients with fewer oncogenic events display lower rates of leukemic transformation and better overall survival, whereas patients with higher number of mutations have a significantly worse outcome (p<0.0001). We evaluate the prognostic potential of targeted genomic sequencing data by means of the area under the receiver operating characteristic (AUC) and compare the predictive value to that of conventional markers utilised in the clinic, including the IPSS classification, peripheral blood counts, bone marrow morphology and cytogenetic analysis. Strikingly we find that the prognostic information contained within the gene sequencing data and peripheral blood counts at diagnosis appears to be equivalent to that derived from existing outcome indicators, suggestive of a potential redundancy between the molecular markers and the morphological variables used to classify patients. Most importantly the number of oncogenic point mutations is an independent biomarker and may therefore be used to refine the prognosis based on existing IPSS categories.

The present study provides an unprecedented characterisation of the genomic lesions implicated in MDS pathogenesis, refines the classification of oncogenic driver mutations and reveals a diverse molecular landscape of mostly rare mutations with valuable insights on the operating biological networks. The absence of known oncogenic mutations in 22% of MDS indicates that additional yet uncharacterised genomic events are likely to be operating in MDS pathogenesis. Most importantly our data highlight the potential of well-defined gene mutations as reproducible diagnostic and predictive biomarkers to support MDS patient management and clinical decision-making in the future.